Articles tagged with "quadruped-robot"
Robot learns adaptive walking on uneven terrain using deep learning
The article discusses a quadruped robot that has successfully learned to walk adaptively on slippery and uneven terrain solely through deep reinforcement learning in simulation, without relying on human-designed gaits or manual tuning. Traditional legged robot control methods depend on precise physical models and predefined motions, which often fail in unpredictable environments. To overcome these limitations, the researchers developed a structured training framework using a curriculum that gradually increases terrain difficulty—from flat ground to slopes, rough surfaces, low friction, and mixed conditions with sensor noise. This approach enables the robot to develop robust locomotion skills and adapt instinctively to new terrains. The robot’s control system features a hierarchical structure with a high-level neural network policy generating joint targets at 10 Hz, executed by a low-level controller at 100 Hz for stability. It uses proprioceptive inputs (joint angles, velocities, body orientation) and exteroceptive data from a simulated depth camera to perceive terrain features. Training employed proximal policy optimization with a multi-objective reward balancing
roboticsdeep-learningadaptive-walkingreinforcement-learningquadruped-robotterrain-navigationrobotic-control-systemsChina's wheeled robot dog climbs stairs at 5 feet per second in demo
Pudu Robotics recently released a video showcasing its PUDU D5 wheeled quadruped robot climbing stairs at a speed of 1.5 meters per second (nearly 5 feet per second) in real time, without edits. The robot demonstrates a hybrid locomotion system, seamlessly switching between wheels on flat terrain and legs for stair climbing, enabling efficient navigation of mixed environments with smooth surfaces and sudden elevation changes. This hybrid approach distinguishes the D5 from other quadrupeds that rely solely on legged movement, emphasizing speed and fluidity. Unveiled in December, the PUDU D5 Series includes two configurations: a fully legged version and a wheeled variant optimized for mixed terrain. Designed for autonomous operation in complex outdoor and industrial settings, the D5 integrates powerful onboard computing using NVIDIA’s Orin platform and an RK3588 chip, supporting real-time mapping, obstacle avoidance, and path planning without constant human supervision. Its 360-degree perception system combines fisheye
robotquadruped-robotautonomous-navigationhybrid-locomotionindustrial-roboticsAI-computingLiDAR-sensorsChina debuts robot dog that can map 10 million square feet nonstop
Chinese robotics company Pudu introduced its latest quadruped robot, the D5, at Tokyo’s International Robot Exhibition (iREX 2025). Standing nearly one meter tall, the D5 showcases advanced motion-control algorithms and embodied intelligence, enabling it to navigate complex environments autonomously. Powered by an NVIDIA Orin platform and RK3588 dual-processor architecture, the robot delivers up to 275 TOPS of computing power for real-time SLAM mapping, obstacle avoidance, and object recognition. It can continuously map and inspect up to one million square meters (approximately 10 million square feet) and travel up to 14 kilometers without human intervention. Equipped with fisheye cameras and LiDAR sensors, the D5 provides 360-degree perception and dense 3D point clouds, enhancing operational safety and efficiency. Designed for durability, it supports a 30-kilogram load for over two hours and is resistant to dust, water, and extreme temperatures. Pudu positions the D5 as
robotautonomous-robotsquadruped-robotSLAM-mappingLiDARNVIDIA-Orinindustrial-robotsUS robot dog patrols massive construction sites for faster progress
FieldAI, an AI robotics company, has partnered with DPR Construction, a major U.S. contractor, to automate construction site processes using autonomous quadruped robots equipped with FieldAI’s autonomy software, Field Foundation Models™. Deployed at a DPR jobsite in Santa Clara, California, the Boston Dynamics Spot robot autonomously conducted extensive surveys and data collection, capturing over 45,000 photos, walking more than 100 miles, and mapping large interior and roofing areas. This system addresses inefficiencies in manual documentation, labor shortages, safety hazards, and operational delays by generating real-time, structured digital records that support decision-making, risk detection, and long-term project documentation. Traditionally, construction documentation involves engineers manually capturing 360° photos over days or weeks, resulting in outdated data and slow progress. FieldAI’s robot navigates dynamic, GPS-free environments autonomously, adapting to daily site changes and performing tasks such as progress tracking, hazard detection, material movement monitoring, and security checks. DPR
roboticsconstruction-automationquadruped-robotAI-roboticsautonomous-navigationconstruction-site-monitoringlabor-shortage-solutionsMeet Sand Hound — The Robot Built to Walk Where Humans Can’t
Sand Hound is a quadruped robot developed collaboratively by the U.S. Army Engineer Research and Development Center and the University of Delaware, designed to navigate challenging coastal terrains such as beaches, dunes, and shifting shorelines where humans and traditional machines often fail. Equipped with advanced sensors including LiDAR and cameras, Sand Hound autonomously maps ground instability, erosion, and terrain changes in real time, adapting to the dynamic and unpredictable nature of sandy environments influenced by tides, storms, and wind. Weighing about 70 pounds and standing roughly two feet tall, Sand Hound combines rugged military-grade durability with sophisticated environmental awareness, enabling it to traverse difficult landscapes without sinking or slipping. Tested along North Carolina’s coastal ranges, the robot demonstrates potential as a vital tool for coastal defense, providing continuous monitoring and data collection that could enhance erosion management and disaster response. Sand Hound represents a new class of autonomous coastal guardians, capable of performing tasks that are dangerous or impossible for humans and conventional vehicles.
robotautonomous-robotquadruped-robotmilitary-technologyterrain-mappingLiDARenvironmental-sensorsWorld-first convertible robot switches between biped, quadruped forms
Direct Drive Technology, a Hong Kong-based robotics firm, has introduced the world’s first fully modular embodied intelligence robot called the D1. This innovative robot can autonomously reconfigure itself to switch between bipedal and quadrupedal forms, allowing it to adapt to various terrains and tasks. The quadruped mode offers stability and is suited for uneven or chaotic terrain, while the biped mode is lighter and more efficient on flat surfaces. The D1 demonstrates versatile capabilities such as rolling over smooth terrain for scouting, walking on rough terrain while carrying payloads up to 100 kg, and even recovering from falls with precision. Each biped section weighs 24.3 kg, can roll at speeds up to 11 km/h, and operates for over five hours on a two-hour charge, powered by a Jetson Orin NX 8GB processor running Ubuntu 22.04. The D1’s modular design allows two biped units to combine into a quadruped, expanding its functional
robotmodular-robotbiped-robotquadruped-robotautonomous-robotDirect-Drive-Technologyembodied-intelligence‘World’s first’ war-ready robot dog that fires grenades unveiled
Skyborne Technologies has unveiled CODiAQ (Controller-Operated Direct-Action Quadruped), described as one of the world’s first war-ready robot dogs capable of autonomously firing grenades. Funded by the U.S. Department of Defense’s Office of the Assistant Secretary for Special Operations and Low-Intensity Conflict, CODiAQ is designed to provide small military units with a remotely operated lethal option that can be rapidly deployed and controlled by a single operator. The system integrates modular weapons, including a HAVOC 40mm grenade launcher and a CHAOS 12-gauge shotgun, supported by advanced AI-assisted targeting software that enables autonomous aiming, target tracking, and engagement in complex environments. CODiAQ is engineered for rugged operational conditions, featuring IP-67 certification for dust and water resistance, and the ability to traverse difficult terrain, climb stairs, and navigate confined spaces. Its AI-driven autonomy allows independent navigation, freeing operators to focus on mission-level decisions. The robot’s precision fire capability supports
robotautonomous-robotmilitary-technologyAI-roboticsrobotic-weaponsquadruped-robotdefense-technologyGhost Robotics’ Vision 60: Soldier’s New Best Friend?
The Ghost Robotics Vision 60 is a quadruped unmanned ground vehicle designed for military applications, resembling a dog but built specifically for combat environments. Weighing 51 kg and equipped with an NVIDIA AI system, it can navigate challenging terrains by climbing, crawling, and swimming—capabilities that surpass traditional wheeled vehicles. Its multifunctional role includes scouting, carrying equipment, and creating 3D threat maps, making it a versatile asset on the battlefield. Engineered to endure extreme conditions ranging from Arctic cold to desert heat, the Vision 60 combines endurance, autonomy, and adaptability to support soldiers in diverse environments. By integrating advanced AI and robust mobility, this robotic platform aims to enhance battlefield reconnaissance and operational efficiency, potentially becoming a critical tool for future military operations.
robotunmanned-ground-vehicleAI-roboticsmilitary-technologyquadruped-robotautonomous-robotbattlefield-roboticsBoston Dynamics’ robot dog nails daring backflips in new video
Boston Dynamics has showcased its robot dog, Spot, performing consistent backflips in a new video, highlighting the robot’s advanced agility and refined design. While these gymnastic feats are unlikely to be part of Spot’s routine tasks, they serve a critical engineering purpose: pushing the robot to its physical limits to identify and address potential balance failures. This helps improve Spot’s ability to recover quickly from slips or trips, especially when carrying heavy payloads in industrial settings, thereby enhancing its reliability and durability. The development of Spot’s backflip capability involved reinforcement learning techniques, where the robot was trained in simulations to optimize its movements by receiving rewards for successful actions, akin to training a dog with treats. This iterative process of simulation and real-world testing allows engineers to fine-tune Spot’s behavior and ensure robust performance. Beyond technological advancements, Spot’s agility has also been demonstrated in entertainment contexts, such as performing dance routines on America’s Got Talent, showcasing its versatility. Looking forward, Spot’s ongoing evolution through
robotroboticsBoston-Dynamicsrobot-dogreinforcement-learningmachine-learningquadruped-robotDoggy Robot Plays Badminton
The article introduces ANYmal, a quadruped robot developed by ETH Zurich primarily for detecting gas leaks in challenging environments. Despite its technical purpose, the robot has been showcased playing badminton, highlighting its agility and advanced mobility. This demonstration serves to illustrate the robot's precise movement capabilities and adaptability beyond industrial applications. By engaging in a dynamic sport like badminton, ANYmal exemplifies the potential for quadruped robots to perform complex, coordinated tasks requiring balance, speed, and responsiveness. The badminton example underscores the progress in robotics that enables machines to operate in diverse scenarios, from safety inspections to interactive activities. However, the article does not provide detailed information on the technical modifications or programming that allow ANYmal to play badminton specifically.
robotquadruped-robotANYmalroboticsgas-leak-detectionETH-Zurichbadminton-robotVideo: Swiss robot dog plays perfect badminton match with a human
Researchers at Switzerland’s ETH Zurich have developed a quadruped robot dog named ANYmal, capable of playing badminton with a human at the skill level of a seven-year-old child. ANYmal, created by ANYbotics, uses a sophisticated control system equipped with two cameras to track and predict the shuttlecock’s trajectory. It swings a racket attached to a multi-axis arm to hit the shuttlecock precisely. The robot was trained using reinforcement learning in a virtual environment, where it practiced thousands of rallies to learn positioning, shot accuracy, and anticipatory movement, enabling it to perform with remarkable precision in real-world play. A key challenge addressed in the development was maintaining balance while lunging and moving quickly to return shots. ANYmal’s reinforcement learning algorithm enhances its coordination and stability, allowing it to move with agility and balance comparable to a human player. Originally designed for industrial inspection and navigating rough terrains, including disaster zones, ANYmal’s capabilities have now been extended to dynamic sports environments. Priced at around
robotroboticsreinforcement-learningquadruped-robotrobot-dogautonomous-robotsrobot-control-systemsChina’s robot dog sprints 328 feet in 16.33 seconds, breaks record
China’s Zhejiang University announced that its quadruped robot, White Rhino, set a new Guinness World Record by sprinting 100 meters (328 feet) in 16.33 seconds, surpassing the previous record of 19.87 seconds held by South Korea’s Hound robot. The run took place in Hangzhou and marks a significant advancement in robotic speed, narrowing the gap between machine and human sprint performance (Usain Bolt’s human record is 9.58 seconds). This achievement demonstrates the robot’s explosive power, speed, stability, and precise control during rapid movement. White Rhino was developed through a collaborative effort involving Zhejiang University’s Center for X-Mechanics, School of Aeronautics and Astronautics, and the Hangzhou Global Scientific and Technological Innovation Center. The design process employed a “robot forward design” approach, using comprehensive dynamics simulations and multi-objective optimization to simultaneously refine geometry, motor specifications, and reduction systems. The robot features high-power-density joint actuators
robotquadruped-robotroboticsactuatorscontrol-algorithmsreinforcement-learningmechanical-designUnitree launches A2 quadruped equipped with front and rear lidar - The Robot Report
Unitree Robotics has launched its latest quadruped robot, the Unitree A2, designed for industrial applications such as inspection, logistics, and research. The A2 features significant upgrades in perception, including dual industrial lidar sensors positioned at the front and rear, an HD camera, and a front light to improve environmental detection and eliminate blind spots. Weighing 37 kg unloaded, the A2 can carry a 25 kg payload while walking continuously for three hours or about 12.5 km, supported by hot-swappable dual batteries for extended missions. This model balances endurance, strength, speed, and perception, marking it as one of Unitree’s most advanced quadrupeds to date. Key specifications of the A2 include a top speed of 5 m/s, an unloaded range of 20 km, a maximum standing load of 100 kg, and the ability to climb steps up to 1 meter high. Compared to Unitree’s previous B2 model, the A2 is
robotquadruped-robotlidarautonomous-robotsroboticsAI-visionbattery-technologyUnitree Releases World's Fastest Quadruped Robot
The article announces Unitree's latest innovation in robotics, the Unitree A2 Stellar Explorer, which is touted as the world's fastest quadruped robot. Following the success of its predecessor, the Unitree R1, the A2 Stellar Explorer represents a significant advancement in speed and agility for four-legged robots. Although specific performance metrics and technical details are not provided in the excerpt, the emphasis is on the robot's enhanced capabilities and potential applications. Unitree continues to push the boundaries of robotic design, focusing on creating agile, dog-like robots that can navigate diverse environments quickly and efficiently. The A2 Stellar Explorer is positioned as a cutting-edge development in this field, likely aimed at industries requiring rapid and versatile robotic mobility. Further details on its features, use cases, and technological innovations would provide a clearer picture of its impact and significance.
robotquadruped-robotUnitreerobotics-technologyautonomous-robotsrobot-innovationrobotic-explorationUnitree’s glass-shattering robot dog scales slopes, carries loads
Unitree Robotics has unveiled its latest quadruped robot dog, the A2, designed for demanding industrial applications with enhanced mobility, endurance, and performance. Weighing about 82 pounds (37 kg), the A2 features 12 degrees of freedom and powerful motors delivering up to 180 Nm of torque, enabling it to carry loads up to 55 pounds (25 kg) and support standing loads of 220 pounds (100 kg). The robot can navigate challenging terrain, including climbing 45° slopes, ascending 30 cm stairs, and traversing rough pathways with agility. Equipped with front and rear industrial-grade LiDAR sensors, an HD camera, and a front light, the A2 can detect and respond to its environment in real time, ensuring precise movement and stability. The A2 demonstrates remarkable agility and durability, as showcased in a promotional video where it performs backflips, balances on one leg, and even crashes through glass without losing functionality. Its 12 high-density motors allow
robotquadruped-robotindustrial-robotLiDARrobot-dogroboticsautonomous-navigationChina’s latest robot dog performs flips, handstands, tough climbs
China’s robotic startup MagicLab has unveiled the MagicDog-W, a wheel-legged quadruped robot that combines wheels and legs to achieve high mobility and agility. Equipped with 17 motors driving each joint and wheel, the robot can perform dynamic stunts such as flips and handstands while navigating challenging terrains including slopes steeper than 40 degrees, stairs, and vertical obstacles up to 60 cm. It can carry a payload of up to 10 kg (22 lbs) and reach speeds of 3 meters per second (6.71 mph), with an operational runtime of 2 to 4 hours per charge. The MagicDog-W’s advanced motor control system allows it to adapt its posture dynamically in response to terrain changes, enabling stable movement on rough and unstructured surfaces. Its combination of speed, endurance, and payload capacity makes it suitable for practical applications such as industrial inspection, search and rescue, exploration, and military logistics. While pricing has not been disclosed, the robot’s innovative wheel-leg
robotroboticsquadruped-robotmotor-controlterrain-navigationrobot-dogagile-roboticsNew quadruped robot climbs vertically 50 times faster than rivals
Researchers at the University of Tokyo’s Jouhou System Kougaka Laboratory (JSK) have developed KLEIYN, a quadruped robot capable of climbing vertical walls up to 50 times faster than previous robots. Unlike other climbing robots that rely on grippers or claws, KLEIYN uses a chimney climbing technique, pressing its feet against two opposing walls for support. Its flexible waist joint allows adaptation to varying wall widths, particularly narrow gaps. The robot weighs about 40 pounds (18 kg), measures 2.5 feet (76 cm) in length, and features 13 joints powered by quasi-direct-drive motors for precise movement. KLEIYN’s climbing ability is enhanced through machine learning, specifically Reinforcement Learning combined with a novel Contact-Guided Curriculum Learning method, enabling it to transition smoothly from flat terrain to vertical surfaces. In tests, KLEIYN successfully climbed walls spaced between 31.5 inches (80 cm) and 39.4 inches (
robotquadruped-robotmachine-learningreinforcement-learningclimbing-robotrobotics-innovationautonomous-robotsRobot dog walks on tough terrain with two legs, withstands kicks
Researchers at the University of Hong Kong’s ArcLab have developed a quadruped robot capable of walking on two legs using a bio-inspired controller called TumblerNet, powered by Deep Reinforcement Learning. This system mimics human balance by integrating estimators for the robot’s center of mass and center of pressure into a closed-loop control, enabling seamless transitions between quadrupedal and bipedal locomotion. The robot can respond to various movement commands, including turning and walking in circles, demonstrating advanced adaptability. The robot’s robustness is notable, as it maintains balance on challenging terrains such as foam pads, loose rocks, sand, and even a beach environment. It withstands external disturbances like pushes and kicks without requiring a separate recovery model, and it can automatically recover from falls caused by obstacles. These capabilities highlight the potential advantages of bipedal robots over traditional quadrupeds, especially for navigating human environments and performing complex tasks in caregiving, disaster response, and human-robot collaboration. The researchers
robotquadruped-robotbipedal-locomotionbio-inspired-controllerdeep-reinforcement-learningrobot-stabilityautonomous-robotsStanford students build tiny AI-powered robot dog from basic kit
Stanford University’s Computer Science 123 course offers undergraduates a hands-on introduction to robotics and AI by having them build and program a low-cost quadruped robot called “Pupper.” Over a 10-week elective, student teams receive a basic robot kit and learn to engineer the platform’s movement, sensing, and intelligence from the ground up. By the course’s end, groups demonstrated Puppers capable of navigating mazes, acting as tour guides, or simulating firefighting with a toy water cannon, showcasing practical applications of their AI and hardware skills. The course originated from a student robotics club project called “Doggo,” designed to prove that advanced legged robots need not be prohibitively expensive. Led by instructors including former Tesla executive Stuart Bowers, Stanford professor Karen Liu, and Google DeepMind researcher Jie Tan, the curriculum guides students from basic motor control and sensor calibration to training neural networks for gait refinement, object tracking, and voice command response. Students even create custom hardware extensions, bridging
robotAIrobotics-educationquadruped-robotStanford-Universityneural-networkshardware-developmentDuke's robot dog mimics human touch, sound to navigate forest terrain
robotAInavigationsensory-technologyquadruped-robotWildFusionrobotics